AI Evaluation & Red-Teaming for Next-Gen AI Agents — HORIZON-CL4-2026-05-DIGITAL-EMERGING-02

Giskard is a French SME and the developer of an open-source AI evaluation and red-teaming platform used across Europe. We are seeking to join a consortium for HORIZON-CL4-2026-05-DIGITAL-EMERGING-02: Next-Generation AI Agents for Real-World Applications.

What we bring to a consortium:

• Automated evaluation and benchmarking of autonomous AI agents (planning, tool use, action execution)

• Red-teaming methodologies for multi-agent systems: detecting cascading failures, adversarial vulnerabilities, and emergent unsafe behaviors

• Hallucination detection, prompt injection resistance testing, and bias assessment for LLM-based agents

• AI Act conformity assessment tooling for high-risk autonomous systems

• An established open-source platform (GitHub: 4k+ stars) with enterprise-grade capabilities

Role we are looking for: Partner responsible for evaluation, benchmarking, and/or trustworthy AI workpackages. We can also contribute to dissemination and standardization activities.

What we are looking for in a consortium: A coordinator with strong expertise in AI agent architectures, and partners bringing real-world application domains in Apply AI sectors (health, manufacturing, mobility).